from typing import Any, Callable, Literal, Optional, Union
import warnings
from segmentation_models_pytorch.base import (
ClassificationHead,
SegmentationHead,
SegmentationModel,
)
from segmentation_models_pytorch.encoders import get_encoder
from segmentation_models_pytorch.base.hub_mixin import supports_config_loading
from .decoder import PANDecoder
[docs]
class PAN(SegmentationModel):
"""Implementation of PAN_ (Pyramid Attention Network).
Note:
Currently works with shape of input tensor >= [B x C x 128 x 128] for pytorch <= 1.1.0
and with shape of input tensor >= [B x C x 256 x 256] for pytorch == 1.3.1
Args:
encoder_name: Name of the classification model that will be used as an encoder (a.k.a backbone)
to extract features of different spatial resolution
encoder_depth: A number of stages used in encoder in range [3, 5]. Each stage generate features
two times smaller in spatial dimensions than previous one (e.g. for depth 0 we will have features
with shapes [(N, C, H, W),], for depth 1 - [(N, C, H, W), (N, C, H // 2, W // 2)] and so on).
Default is 5
encoder_weights: One of **None** (random initialization), **"imagenet"** (pre-training on ImageNet) and
other pretrained weights (see table with available weights for each encoder_name)
encoder_output_stride: 16 or 32, if 16 use dilation in encoder last layer.
Doesn't work with ***ception***, **vgg***, **densenet*`** backbones.Default is 16.
decoder_channels: A number of convolution layer filters in decoder blocks
decoder_interpolation: Interpolation mode used in decoder of the model. Available options are
**"nearest"**, **"bilinear"**, **"bicubic"**, **"area"**, **"nearest-exact"**. Default is **"bilinear"**.
in_channels: A number of input channels for the model, default is 3 (RGB images)
classes: A number of classes for output mask (or you can think as a number of channels of output mask)
activation: An activation function to apply after the final convolution layer.
Available options are **"sigmoid"**, **"softmax"**, **"logsoftmax"**, **"tanh"**, **"identity"**,
**callable** and **None**. Default is **None**.
upsampling: Final upsampling factor. Default is 4 to preserve input-output spatial shape identity
aux_params: Dictionary with parameters of the auxiliary output (classification head). Auxiliary output is build
on top of encoder if **aux_params** is not **None** (default). Supported params:
- classes (int): A number of classes
- pooling (str): One of "max", "avg". Default is "avg"
- dropout (float): Dropout factor in [0, 1)
- activation (str): An activation function to apply "sigmoid"/"softmax"
(could be **None** to return logits)
kwargs: Arguments passed to the encoder class ``__init__()`` function. Applies only to ``timm`` models. Keys with ``None`` values are pruned before passing.
Returns:
``torch.nn.Module``: **PAN**
.. _PAN:
https://arxiv.org/abs/1805.10180
"""
@supports_config_loading
def __init__(
self,
encoder_name: str = "resnet34",
encoder_depth: Literal[3, 4, 5] = 5,
encoder_weights: Optional[str] = "imagenet",
encoder_output_stride: Literal[16, 32] = 16,
decoder_channels: int = 32,
decoder_interpolation: str = "bilinear",
in_channels: int = 3,
classes: int = 1,
activation: Optional[Union[str, Callable]] = None,
upsampling: int = 4,
aux_params: Optional[dict] = None,
**kwargs: dict[str, Any],
):
super().__init__()
if encoder_output_stride not in [16, 32]:
raise ValueError(
"PAN support output stride 16 or 32, got {}".format(
encoder_output_stride
)
)
upscale_mode = kwargs.pop("upscale_mode", None)
if upscale_mode is not None:
warnings.warn(
"The usage of upscale_mode is deprecated. Please modify your code for decoder_interpolation",
DeprecationWarning,
stacklevel=2,
)
decoder_interpolation = upscale_mode
self.encoder = get_encoder(
encoder_name,
in_channels=in_channels,
depth=encoder_depth,
weights=encoder_weights,
output_stride=encoder_output_stride,
**kwargs,
)
self.decoder = PANDecoder(
encoder_channels=self.encoder.out_channels,
encoder_depth=encoder_depth,
decoder_channels=decoder_channels,
interpolation_mode=decoder_interpolation,
)
self.segmentation_head = SegmentationHead(
in_channels=decoder_channels,
out_channels=classes,
activation=activation,
kernel_size=3,
upsampling=upsampling,
)
if aux_params is not None:
self.classification_head = ClassificationHead(
in_channels=self.encoder.out_channels[-1], **aux_params
)
else:
self.classification_head = None
self.name = "pan-{}".format(encoder_name)
self.initialize()